The Inventory Forecasting module estimates future replenishment needs by projecting inventory forward for each product and forecast period based on a seasonally adjusted exponentially weighted moving-average demand forecast model.

This article explains the algorithm using formulas only.

Let:

• P = selected number of forecast periods

• L = selected forecast period length

• H = selected history range in months

• D = selected demand decay factor (monthly)

• ADM = average days per month (30.4375)

For each future period Pₙ, the algorithm defines:

• PeriodStartₙ

• PeriodEndₙ (exclusive)

• PeriodDaysₙ = Days(PeriodStartₙ, PeriodEndₙ)

• ForecastPeriodMonthsₙ = PeriodDaysₙ / ADM (display only)

The history window covers:

• HistoryStartDate = EndDateExclusive − H months

• HistoryDayCount = Days(HistoryStartDate, EndDateExclusive)

For each product and warehouse, historical demand is grouped by calendar day:

• DailyQtyd = total historical demand on day d

The monthly decay factor D is converted to a daily equivalent:

• DailyDecayFactor = D^(1 / ADM)

Each historical day is weighted by recency:

• Weightd = DailyDecayFactor^(DaysAgod − 1)

where DaysAgod = Days(d, EndDateExclusive).

Weighted daily quantity:

• WeightedDailyQty = Σ(DailyQtyd × Weightd)

Weight total across the full history window:

• WeightTotal = (1 − DailyDecayFactor^HistoryDayCount) / (1 − DailyDecayFactor)

• (= HistoryDayCount when DailyDecayFactor = 1)

Weighted daily demand rate:

• WeightedDailyDemand = WeightedDailyQty / WeightTotal (units / day)

Overall average monthly demand (used only as the seasonality baseline):

• OverallAverageMonthlyDemand = (Σ DailyQtyd / HistoryDayCount) × ADM

For each future period, the algorithm slices the period into calendar-month segments and computes a demand-weighted seasonal average.

For each calendar month M overlapped by period Pₙ:

• SeasonalAverageForMonthM = average of HistoricalSalesByCalendarMonth for month M, starting from the item's first sale date

• DaysInSlice = days that period Pₙ falls within month M

Period-level seasonal average monthly demand:

• SeasonalAverageMonthlyDemandₙ = Σ(SeasonalAverageForMonthM × DaysInSlice) / Σ(DaysInSlice)

Seasonality factor:

• SeasonalityFactorₙ = SeasonalAverageMonthlyDemandₙ / OverallAverageMonthlyDemand

• (= 1 when OverallAverageMonthlyDemand ≤ 0)

Base forecast before seasonality (daily rate × period days):

• ForecastDemandBeforeSeasonalityₙ = WeightedDailyDemand × PeriodDaysₙ

Forecast after seasonality before rounding:

• ForecastDemandAfterSeasonalityBeforeRoundingₙ = WeightedDailyDemand × SeasonalityFactorₙ × PeriodDaysₙ

Final forecast demand:

• ForecastDemandₙ = ROUND(ForecastDemandAfterSeasonalityBeforeRoundingₙ, 0)

For each period:

• NonShippedSOQtyₙ = approved sales order demand expected in period n

• NonAllocatedWOQtyₙ = production-related demand expected in period n

• NonReceivedPOQtyₙ = approved purchase order supply expected in period n

• NonReceivedWOQtyₙ = approved production supply expected in period n

The expected date is set in the respective Purchase/Sales Orders in the Latest Expected Date field. For the Production Orders, this is stored under the Due Date field.

For each period:

• DemandToCoverₙ = ForecastDemandₙ + NonShippedSOQtyₙ + NonAllocatedWOQtyₙ

For each product and period, the algorithm aggregates across all selected inventory locations:

• CurrentStockLevelₙ = Σ(current stock across selected locations)

• StockAlertQtyₙ = Σ(stock alert level across selected locations)

• ForecastDemandₙ = Σ(forecast demand per location)

• DemandToCoverₙ = Σ(demand to cover per location)

• NonReceivedPOQtyₙ = Σ(incoming PO quantity)

• NonReceivedWOQtyₙ = Σ(incoming WO quantity)

• StartingQty₁ = CurrentStockLevel₁

Projected quantity before planned purchase:

• ProjectedBeforePurchase₁ = StartingQty₁ + NonReceivedPOQty₁ + NonReceivedWOQty₁ − DemandToCover₁

Planned purchase quantity:

• PlannedPurchaseQty₁ = MAX(0, StockAlertQty₁ − ProjectedBeforePurchase₁)

Projected quantity after planned purchase:

• ProjectedAfterPurchase₁ = ProjectedBeforePurchase₁ + PlannedPurchaseQty₁

For each period n > 1:

• StartingQtyₙ = ProjectedAfterPurchaseₙ₋₁

Projected quantity before planned purchase:

• ProjectedBeforePurchaseₙ = StartingQtyₙ + NonReceivedPOQtyₙ + NonReceivedWOQtyₙ − DemandToCoverₙ

Planned purchase quantity:

• PlannedPurchaseQtyₙ = MAX(0, StockAlertQtyₙ − ProjectedBeforePurchaseₙ)

Projected quantity after planned purchase:

• ProjectedAfterPurchaseₙ = ProjectedBeforePurchaseₙ + PlannedPurchaseQtyₙ

For each product and period, the module displays:

• Forecast Demand = ForecastDemandₙ

• Demand to Cover = DemandToCoverₙ

• Starting Quantity = StartingQtyₙ

• Projected Quantity = ProjectedBeforePurchaseₙ

• Total Stock Alert Level = StockAlertQtyₙ

• Planned Purchase Quantity Needed = PlannedPurchaseQtyₙ

For each period n:

• ForecastDemandₙ = ROUND(WeightedDailyDemand × SeasonalityFactorₙ × PeriodDaysₙ, 0)

• DemandToCoverₙ = ForecastDemandₙ + NonShippedSOQtyₙ + NonAllocatedWOQtyₙ

• ProjectedBeforePurchaseₙ = StartingQtyₙ + NonReceivedPOQtyₙ + NonReceivedWOQtyₙ − DemandToCoverₙ

• PlannedPurchaseQtyₙ = MAX(0, StockAlertQtyₙ − ProjectedBeforePurchaseₙ)

• ProjectedAfterPurchaseₙ = ProjectedBeforePurchaseₙ + PlannedPurchaseQtyₙ

And for n > 1:

• StartingQtyₙ = ProjectedAfterPurchaseₙ₋₁